• Results
• Table N.1. Between-Subjects Measurement Invariance by Language Version (French vs. English): CAARS 2 Self-Report
• Table N.2. Between-Subjects Measurement Invariance by Language Version (French vs. English): CAARS 2 Observer
• Table N.3. Between-Subjects Measurement Invariance by Language Version (Spanish vs. English): CAARS 2 Self-Report
• Table N.4. Between-Subjects Measurement Invariance by Language Version (Spanish vs. English): CAARS 2 Observer
• Table N.5. Differential Test Functioning by Language Version (French vs. English)
• Table N.6. Differential Test Functioning by Language Version (Spanish vs. English)

In addition to the within-subject invariance approach outlined in Content and DSM Symptom Scales in chapter 13, Translations, a between-subjects approach was taken to examine whether ratings from the French (Canada) or Spanish (North American) language versions of the Conners Adult ADHD Rating Scale 2nd Edition (CAARS™ 2) differed from a demographically matched sample taken from the English Normative Sample (individuals were matched on gender, age, race/ethnicity, and education level). Both a multiple-group confirmatory factor analysis (CFA) and an item-response theory (IRT; specifically differential test functioning [DTF]) approach were taken for the between-subject sample. Details on the methodology for these analyses are covered in appendix M, Methods of Evaluating Measurement Bias.

Results

Results of the between-subjects MI analyses for the French and Spanish translation are presented in Table N.1 to Table N.4. Overall, the models were found to be invariant between the French and English and Spanish and English versions of the forms, as evidenced by non-decreasing ΔCFI values and nonsignificant Satorra-Bentler chi-square tests. As part of the modeling procedure, some steps required partial invariance adjustments to result in nonsignificant chi-square tests, but the adjustments were infrequent and did not compromise the overall comparability of the scales between the language versions (Dimitrov, 2010). The results reflect the findings from the within-subject MI analysis reported in chapter 13, Translations, and provide additional evidence for the validity of the CAARS 2 French (Canada) and Spanish (North American) translations as parallel measures to the English version.

In addition to the MI analyses, DTF was also examined for Content Scales. Across both the French and English and Spanish and English comparisons, test characteristic curves for Content Scales were found to be statistically equivalent as evidenced by overlapping 95% confidence intervals (see Figure N.1 and Figure N.2 for examples featuring the Inattention/Executive Dysfunction scale). More specifically, considerable overlap occurred in the area approaching and exceeding 1.5 standard deviations above mean theta (see Test Information in chapter 8, Reliability, for more information on how these graphs are interpreted). This pattern of results demonstrates a lack of difference between the functioning of Content Scales for the French and English and Spanish and English language versions, a finding further supported by the negligible effect sizes presented in Table N.5 and Table N.6.

Click to expand

Figure N.1. Differential Test Functioning for French and English Language Versions (Inattention/Executive Dysfunction)

a) Self-Report

b) Observer

Click to expand

Figure N.2. Differential Test Functioning by Spanish and English Language Versions (Inattention/Executive Dysfunction)

a) Self-Report

b) Observer

Both the muliple-group CFA and DTF approach support the invariance of the factor structure for Content Scales on the CAARS 2 French and Spanish versions. This between-subject approach adds support to the within-subject approach presented in chapter 13, Translations.